A group of 26 current and former Meta employees who say the company's artificial-intelligence systems steered them toward the chopping block are pushing their case into its next, more revealing phase. After a federal judge in California declined to block their layoffs but flagged "serious questions" about how Meta made its cuts, the dispute is now moving from opening arguments into the evidence-gathering stage that will determine whether one of the first major AI-driven layoff discrimination claims in the country has legs.
The case, filed July 13 in the U.S. District Court for the Northern District of California, is widely described as the first U.S. lawsuit accusing a major technology employer of using AI to make discriminatory layoff decisions. What happens next — what Meta must disclose about its internal tools, and when — could shape how courts treat algorithmic employment decisions for years.
The allegations
The plaintiffs, proceeding anonymously as Does 1 through 26, allege that Meta relied on what their lawyers call a "constellation of internal artificial-intelligence systems" to score, rank and select employees during a reduction in force that Meta began notifying staff about on May 20. Reporting places the total cut at roughly 10% of Meta's workforce, or on the order of several thousand people.
According to the complaint, that scoring machinery penalized workers who had taken legally protected leave. The plaintiffs contend the systems measured productivity through signals such as keystroke and activity monitoring, AI "token" consumption dashboards, and performance ratings tied in part to employees' adoption of AI tools — including an internal assistant reported to be called Metamate. Because those metrics allegedly could not account for approved absences, the suit argues, employees who were out on maternity, medical, parental or caregiving leave showed up as low performers through no fault of their own.
The composition of the plaintiff group is central to the theory. Per court filings described in news reports, eight are women who took maternity or pregnancy-related leave, four are men who took parental leave, and at least one is a woman who took leave to care for a family member and later took bereavement leave. Plaintiffs' lawyers argue that a selection process that "systematically record[s] such absences as reduced performance" falls "more heavily on women than on men."
The suit invokes several federal statutes, including the Family and Medical Leave Act, the Americans with Disabilities Act, the Pregnancy Discrimination Act and the Pregnant Workers Fairness Act.
These allegations are unproven. Meta denies them. "These claims lack merit and are not based on facts," a company spokesperson said, adding: "Workforce management and organizational decisions were and are made by people, not AI."
Where the case stands
A note of caution on the "discovery" framing: this dispute is procedurally unusual. The plaintiffs have said they wish to pursue their claims individually in private arbitration, and U.S. District Judge William Orrick, an Obama appointee, has indicated the merits will be decided there rather than in open court. So the traditional, court-supervised discovery of a class action is not quite what is unfolding.
What is unfolding is the beginning of a fight over evidence. On July 17, Judge Orrick denied the workers' request for a temporary restraining order that would have paused their separations, set to begin July 22, finding they had not shown the "irreparable harm" needed for emergency relief. But his written order acknowledged the case raises "serious questions going to the merits," and he left the door open to revisiting the decision. In a joint statement, plaintiffs' lawyers noted that "the Court expressly stated that it may reconsider its determinations 'based on any additional evidence the parties provide regarding whether and how AI was used' in the reduction in force."
Crucially, the court directed Meta to explain — by July 23 — how and why four visa-dependent plaintiffs were selected for termination. Legal observers described that as the first time Meta will be required to account for any specific selection decision. That order, and the judge's invitation for more evidence about the role of AI, is the practical on-ramp to discovery: the point at which Meta's internal ranking tools stop being a black box and start becoming a documented record.
Why it matters
The case lands at the center of a fast-moving debate over employer liability for algorithmic decisions. Regulators and plaintiffs' attorneys have increasingly argued that "disparate impact" — a neutral-looking process that disproportionately harms a protected group — can attach to AI systems even when no human intended to discriminate. Meta's defense, that decisions "were and are made by people, not AI," previews the central factual battle: whether AI tools merely informed human managers or effectively drove the outcomes.
If the plaintiffs can show through documents and testimony that Meta's scoring systems functioned as decision-makers and that the company did not audit them for this failure mode, the case could become a template. If Meta shows meaningful human review, it may reassure employers that "human-in-the-loop" processes remain a viable shield.
What to watch
Watch whether Meta's July 23 explanation satisfies the court or prompts Orrick to reconsider — and whether the arbitration track limits how much of the evidentiary record ever becomes public. Watch, too, for how much the company is compelled to reveal about Metamate and its performance-calibration tools. Because so much may play out in private arbitration, the biggest open question is whether this landmark test of employment-AI liability produces a public precedent at all — or a confidential settlement that leaves the law unsettled.
"These claims lack merit and are not based on facts. Workforce management and organizational decisions were and are made by people, not AI."— Meta spokesperson, Company statement